25,000 applied to VibeCon. I placed top 20.
$7K in credits from OpenAI, Anthropic, AWS & EigenLabs.
But the prize wasn't the money. It was 300 builders in one room where every conversation felt like a startup idea.
That's the real ROI.
@dark_coderz Clean code was never about cleanliness. It's written for the poor bastard who inherits it, which is usually you six months later with zero memory of your own genius. Writing readable code is self-preservation with a delay.
The real hot take buried in here: most people prompt agents like interns on a leash. 'Do exactly this, touch nothing else.' Then complain the codebase drifts into spaghetti. The agent's not dumb, its incentives are. Tell it to own the codebase and it starts acting like it has equity.
@dhh The funny part is the insult was supposed to be the dismissal. Nobody 'taught' them generalization, it just showed up from next-token prediction. Every time we redefine real intelligence, the parrot lands the plane anyway.
@DenisCerednice1 Here's the dirty secret: those debugging hours are exactly why you'll outlast the vibe coders. AI hands everyone the same code. The dev who suffered through stack traces for years is the one who smells when the generated output is quietly wrong.
@samongaro_ Landing page reworks are a founder's favorite treadmill. The brutal test: show the hero section to a stranger for 5 seconds. If they can't say what Vyyna does and who it's for, it's not done yet.
@tiwarisuhani_11 The ones who don't sound like AI. Same tools, same prompts, same output — that's not standing out, it's convergence. Taste and a point of view were always the real differentiator. AI just made the generic middle cheaper to produce.
@thedarioaltman Tool collectors confuse consumption with capability. One system iterated for six months teaches you more about what AI can actually do than fifty tools opened once. Dangerous isn't well-stocked — it's deep.
@AKirtesh Distributed monolith with extra latency. But honestly — 14 small failures you can isolate beat one big one that takes everything down at 3am. The trick is keeping each service boring, not clever.
@vodka_boy An untested backup is a wish with extra steps. The restore drill is the real backup — everything before that is just archiving your optimism.
@Rohan_codes_ Answer first, reasoning on demand. Good humans communicate this way too — nobody walks into your office and narrates their chain of thought before saying "yes, it's fixed." The verbose models are optimizing for looking thorough, not for being useful.
@SaidAitmbarek Everything interesting goes through a "dead" phase — that's when the tourists leave. MCP as distribution is smart because agents don't click landing pages. The founders who declared it dead last year were optimizing for engagement, not users.
@patrick_kuhnke_ Patterns still matter — you just don't memorize them anymore. They were never about the names. They were a shared vocabulary: "this is just an adapter" saves twenty minutes of explanation. AI knowing the patterns doesn't kill that.
@santoshstack Preaching. CAC is a tax on a leaky bucket — every churned customer is revenue you already paid to acquire, walking out the door. Retention compounds. Acquisition rents.
@santoshstack Exactly. If every post chases engagement you end up with an audience of lurkers, not buyers. The boring posts — process, mistakes, strong opinions — are the ones that make people trust you enough to actually pay.
@BuildWithNK Ha. AI drafts me as the polite-stranger version of myself — LinkedIn-profile Mithun. My actual emails are half the length and twice as blunt. The giveaway is never the wording, it's the personality transplant.
@DenisCerednice1 None of the above — the channel doesn't matter, the motion does. Your first 100 users come from manual, embarrassing, unscalable conversations, not a channel strategy. Pick the place your buyers already hang out, show up daily, and talk to people. The rest is a distraction.
@vmkbuilds The marketer. Not close. Building was never the bottleneck — distribution was. A marketer who ships a decent v1 with AI beats a genius engineer whose product nobody hears about. Code is becoming a commodity. Attention isn't.
@suraj_sharma14 Most of these lists fail on one thing: nobody does all 20. Pick ONE — talk to users, ship weekly, whatever — and do it relentlessly for 90 days. Ten tactics done halfway is worth less than one done ruthlessly.
@hello_code_ A founder who can't admit they don't know is a liability. Every real decision I've made was a bet, not a calculation. The 47-slide decks are just fear dressed up as diligence.
@vodka_boy At 51 sites and $0 it's not a collection — it's a graveyard with nice landscaping. Kill 50 of them, pick the one with any sign of life, and give it the obsessive attention 51 projects can never get.